search for: crypto_send

This time around, the numbers are indeed looking better than in v1; and I like the interface better. Eric Blake (2): server: Prefer send() over write() server: Group related transmission send()s server/internal.h | 7 +++- server/connections.c | 51 +++++++++++++++++++++++++--- server/crypto.c | 11 ++++--

...there's no need to try and merge it + * with any corked data from a previous send that used SEND_MORE. + */ +#define MAX_SEND_MORE_LEN (64 * 1024) + /* Write buffer to GnuTLS and either succeed completely * (returns 0) or fail (returns -1). flags is ignored for now. */ @@ -357,7 +363,11 @@ crypto_send (struct connection *conn, const void *vbuf, size_t len, int flags) assert (session != NULL); - if (flags & SEND_MORE) + if (len >= MAX_SEND_MORE_LEN) { + if (gnutls_record_uncork (session, GNUTLS_RECORD_WAIT) < 0) + return -1; + } + else if (flags & SEND_MORE) gn...

...return 0; +} + /* Read buffer from conn->sockin and either succeed completely * (returns > 0), read an EOF (returns 0), or fail (returns -1). */ diff --git a/server/crypto.c b/server/crypto.c index 978a843..e4abca2 100644 --- a/server/crypto.c +++ b/server/crypto.c @@ -371,6 +371,24 @@ crypto_send (struct connection *conn, const void *vbuf, size_t len) return 0; } +/* Change the cork status to batch a group of send calls, and either succeed + * completely (returns 0) or fail (returns -1). + */ +static int +crypto_cork (struct connection *conn, bool cork) +{ + gnutls_session_t session =...

...) { - r = send (sock, buf, len, 0); + r = send (sock, buf, len, f); if (r == -1) { if (errno == EINTR || errno == EAGAIN) continue; diff --git a/server/crypto.c b/server/crypto.c index 3f87944..3f3d96b 100644 --- a/server/crypto.c +++ b/server/crypto.c @@ -357,6 +357,9 @@ crypto_send (struct connection *conn, const void *vbuf, size_t len, int flags) assert (session != NULL); + if (flags & SEND_MORE) + gnutls_record_cork (session); + while (len > 0) { r = gnutls_record_send (session, buf, len); if (r < 0) { @@ -368,6 +371,10 @@ crypto_send (struct...

There are a few other issues that Coverity found, but I believe all can be ignored ... except one: We don't set umask anywhere inside nbdkit. Coverity complains that this is a problem where we create temporary files, since the result of mkstemp depends implicitly on the umask value. I think we might consider setting umask anyway (eg. to 022) just to make plugin behaviour more predictable.

Slightly RFC, as I need more time to investigate why Unix sockets appeared to degrade with this patch. But as TCP sockets (over loopback to localhost) and TLS sessions (regardless of underlying Unix or TCP) both showed improvements, this looks like a worthwhile series. Eric Blake (2): server: Add support for corking server: Cork around grouped transmission send()s server/internal.h | 3

These are a couple of patches in preparation for the Block Status implementation. While the patches (especially the second one) are very large they are really just elementary code motion. Rich.

...en) { + struct connection *conn = GET_CONN; gnutls_session_t session = conn->crypto_session; char *buf = vbuf; ssize_t r; @@ -355,8 +356,9 @@ crypto_recv (struct connection *conn, void *vbuf, size_t len) * (returns 0) or fail (returns -1). flags is ignored for now. */ static int -crypto_send (struct connection *conn, const void *vbuf, size_t len, int flags) +crypto_send (const void *vbuf, size_t len, int flags) { + struct connection *conn = GET_CONN; gnutls_session_t session = conn->crypto_session; const char *buf = vbuf; ssize_t r; @@ -392,8 +394,9 @@ crypto_send (struct...

v1 was here: https://www.redhat.com/archives/libguestfs/2020-February/msg00081.html v2 replaces struct connection *conn = GET_CONN; with GET_CONN; which sets conn implicitly and asserts that it is non-NULL. If we actually want to test if conn is non-NULL or behave differently, then you must use threadlocal_get_conn() instead, and some existing uses do that. Rich.

The third patch is a large but mechanical change which gets rid of passing around struct connection * entirely within the server, preferring instead to reference the connection through thread-local storage. I hope this is a gateway to simplifying other parts of the code. Rich.